Changes observed even in an undiagnosed risk group

People carrying GBA1 gene variants who had not been diagnosed with Parkinson’s disease showed gut microbiome composition intermediate between that of healthy controls and patients. Among these carriers, those with larger patient-like microbiome shifts also had more prominent prodromal symptoms that can arise before disease onset. Published in Nature Medicine, the study suggests that the gut microbiome may offer early risk markers. It does not, however, establish how well these markers predict an individual’s future disease onset or validate an early diagnostic test.

GBA1 variants are a major genetic risk factor for Parkinson’s disease, but not all carriers develop the disease. To identify clues that could help distinguish subgroups within this risk population, the researchers combined clinical assessments with stool metagenomic analysis. Metagenomic analysis reads genetic information from the microbes in a sample collectively, providing insight into microbial composition and functional potential.

Intermediate profiles in stool analyses from 464 people

The multicenter cross-sectional study, conducted in the UK and Italy, included 540 people in its clinical analyses. Of these, microbiome analysis was completed for 464: 271 patients with Parkinson’s disease, 43 GBA1 variant carriers without a Parkinson’s diagnosis, and 150 healthy controls without the variants in question. Because this was a cross-sectional study, its findings should be distinguished from results tracking changes in the same people before and after disease onset.

A total of 176 microbial species differed between patients and healthy controls at unadjusted P<0.05; 103 species met q<0.05 after multiple-testing correction. The researchers used Cliff’s δ, a statistic that captures the magnitude and direction of differences between groups, to examine whether each species increased or decreased in the same direction among undiagnosed carriers as among patients.

Of the 176 species, 142 differed in the same direction, with the average magnitude of difference smaller in carriers than in patients. These species accounted for about one-quarter of each group’s microbiome. The abundance and proportion of species that increased in patients were at intermediate levels in carriers, while species that decreased in patients showed the opposite pattern. These changes included a reduction in butyrate-producing bacteria such as Roseburia. This does not mean that all 142 species individually differed significantly between carriers and controls.

Associations with clinical features, but future onset remains untested

Among patients, larger microbiome shifts tended to be associated with longer disease duration and more severe motor and non-motor symptoms. An analysis that distinguished medication dose from disease duration supported the researchers’ interpretation that microbial differences were associated more with disease course than with treatment itself. However, it did not establish changes over time or a causal relationship. Among undiagnosed variant carriers, a patient-like microbiome profile was also linked to the burden of prodromal symptoms, including autonomic dysfunction.

The researchers also developed an exploratory “Parkinson’s disease microbiome score (PDMS-16),” combining the presence or absence of 16 species. Among healthy controls, the 27 people with high scores had more prominent constipation, autonomic dysfunction, depression, and anxiety than the comparison group. This finding distinguished a subgroup with different clinical features. It did not establish that these people developed the disease or validate their probability of developing it in the future.

Replication in independent cohorts and remaining questions

Similar microbiome differences were found in independent cohorts from the US, South Korea, and Türkiye, comprising 638 patients and 319 healthy controls. This result mainly replicated differences between patients and controls. In the Türkiye cohort, where clinical data were available, some associations between patients’ microbiome shifts and clinical features such as disease duration were also replicated. These findings should not be taken as independent validation of future-onset prediction in undiagnosed people or of the predictive performance of PDMS-16.

Key limitations include the cross-sectional design, the small number of undiagnosed variant carriers (43), and the shortage of independent cohorts of undiagnosed people with sufficient clinical information. The study also cannot determine whether gut changes occur before brain pathology, or whether they are a cause or consequence. To establish their value as early markers, sufficiently large groups of people will need to be followed over a prolonged period to determine whether microbial measures are linked to actual disease onset.

Detailed results and methods are available in the Nature Medicine paper and the open-access full text.